Objective measurement of GBCA leakage

The 3D-real IR and MRC images, with anonymized patient information, were imported into Fiji/ImageJ software version 1.53a (NIH, Bethesda, USA). A smooth filter was applied to the 3D-real IR images to exclude small areas of high signal intensity, such as the brain cortex boundaries. These images were then binarized to allow the measurement of leaking GBCA under narrow window conditions (width = 2, level = 30; Fig. 1a). The window level and width parameters were chosen based on a previous study [6]. The MRC images were also binarized to identify the areas where CSF was present under narrow window conditions (width = 2, level = 180; Fig. 1b). Averages (Fig. 1c) were calculated between these two types of binarized images. The slices above the lateral ventricles were extracted to quantify the voxels exhibiting GBCA leakage into the CSF (hereafter named V-leak).

a A 3D-real inversion recovery image binarized to measure Gadolinium-based contrast agent leakage under narrow window conditions (width = 2, level = 30). b A magnetic resonance cisternography image binarized to identify areas where cerebrospinal fluid was present under narrow window conditions (width = 2, level = 180). c Average image between these two types of binarized images to calculate the number of voxels indicating leakage of Gadolinium-based contrast agent into the cerebrospinal fluid